Abstract
This study presents the Transformer framework, which is a top performer in the field of natural language time-series prediction, and suggests an enhanced Transformer neural network model to address the application needs of semi-submersible production platforms for very short-term prediction of mooring tension during operation. Compared with the traditional recurrent neural network, the Transformer neural network has faster computation speed, better generalization ability and scalability, and stronger ability to capture long-range dependencies. In this paper, the proposed Transformer model is used for mooring tension prediction of semi-submersible production platforms, and the proposed neural network model is trained and validated using the actual monitoring data of the first ultra-deepwater semi-submersible production, storage, and offloading platform in the South China Sea during operation. The results show that the proposed method can achieve almost the same prediction results as the current mainstream LSTM model, which has the value of engineering application and promotion.
